[c-nsp] 3750 12.2(44)SE1 CPU 5% weirdness

[Andre Beck]

On Fri, May 16, 2008 at 10:51:03AM +0200, sthaug at nethelp.no wrote:
> > This is normal for those platforms which do almost all that is essential
> > in forwarding traffic using their ASICs, with only initial packets of
> > new flows beeig punted up to the CPU.
> 
> No, the "initial packets of new flows being punted up to the CPU" is
> *not* the case for this platform (or for any modern switch platform).
> It was an issue for older 6500 boxes (Sup1a?).

I stand corrected.

I assumed that to be the case because on the first 3750 that hit my
lab, I configured the (unsupported on that platform) NetFlow data
export and I seem to remember that I have seen *some* traffic, but
it was just a very small trickle, orders of magnitude below the
actual flows going through the box. I assumed that what I see are
just the punts.

Or was this the broken counters on SVIs? Damn LRU brain cache line
dropping...

Probably what I saw was just traffic hitting the control plane?

Thanks,
Andre.
-- 
   Real men don't make backups of their mail. They just send it out
    on the Internet and let the secret services do the hard work.

-> Andre Beck    +++ ABP-RIPE +++      IBH IT-Service GmbH, Dresden <-